Police Shooting

Description

Features of the Kaggle dataset are:

• id: case id
• name: Nomial data for name
• date: Timestamp of the shooting case
• manner_of_death: Victim was either shot or shot and tasered
• armed: Type of armed weapon the victime used
• age: Age of the victim
• gender: Gender of victim M (male) or F (female)
• race: Race of the victim with 6 categories: Asian, Black, White, Native, Hispanic, and Other
• city: City where the victim was shot
• state: State where the victim was shot
• signs_of_mental_illness: True or False if the victim has mental illness
• threat_level: Either the victim attacked, undetermined, or Other
• flee: The victim flet by Foot, Car, Not fleeing, or Other
• body_camera: Either the police had body camera on or not (True or False)
• arms_category: Category of arm that victim had

import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
import seaborn as sns
sns.set_style('whitegrid')
%matplotlib inline
import plotly.graph_objs as go #importing graphical objects
from plotly.offline import download_plotlyjs, init_notebook_mode, plot, iplot
from wordcloud import WordCloud, STOPWORDS, ImageColorGenerator

from sklearn.model_selection import train_test_split
from sklearn.decomposition import PCA
from sklearn.preprocessing import StandardScaler
from sklearn.model_selection import GridSearchCV
from sklearn.linear_model import LogisticRegression
from sklearn.tree import DecisionTreeClassifier
from sklearn.ensemble import RandomForestClassifier
from sklearn.metrics import precision_score, recall_score
from sklearn.metrics import classification_report, confusion_matrix

df = pd.read_csv('shootings.csv')

shooting =df

shooting.describe()

	id	age
count	4895.000000	4895.000000
mean	2902.148519	36.549750
std	1683.467910	12.694348
min	3.000000	6.000000
25%	1441.500000	27.000000
50%	2847.000000	35.000000
75%	4352.500000	45.000000
max	5925.000000	91.000000

shooting.groupby(['race']).size()

race
Asian         93
Black       1298
Hispanic     902
Native        78
Other         48
White       2476
dtype: int64

shooting['shooting count'] =  shooting['name'].groupby(shooting['state']).transform('count')

shooting.head(5)

	id	name	date	manner_of_death	armed	age	gender	race	city	state	signs_of_mental_illness	threat_level	flee	body_camera	arms_category	shooting count
0	3	Tim Elliot	2015-01-02	shot	gun	53.0	M	Asian	Shelton	WA	True	attack	Not fleeing	False	Guns	126
1	4	Lewis Lee Lembke	2015-01-02	shot	gun	47.0	M	White	Aloha	OR	False	attack	Not fleeing	False	Guns	76
2	5	John Paul Quintero	2015-01-03	shot and Tasered	unarmed	23.0	M	Hispanic	Wichita	KS	False	other	Not fleeing	False	Unarmed	49
3	8	Matthew Hoffman	2015-01-04	shot	toy weapon	32.0	M	White	San Francisco	CA	True	attack	Not fleeing	False	Other unusual objects	701
4	9	Michael Rodriguez	2015-01-04	shot	nail gun	39.0	M	Hispanic	Evans	CO	False	attack	Not fleeing	False	Piercing objects	168

shooting.groupby('flee').size()

flee
Car             820
Foot            642
Not fleeing    3073
Other           360
dtype: int64

state_shooting= shooting['shooting count'].groupby(shooting['state']).value_counts()
state_shooting.sort_values(ascending = False)

state  shooting count
CA     701               701
TX     426               426
FL     324               324
AZ     222               222
CO     168               168
GA     161               161
OK     151               151
NC     148               148
OH     146               146
WA     126               126
TN     125               125
MO     124               124
LA     102               102
IL     99                 99
AL     95                 95
PA     95                 95
NM     93                 93
VA     92                 92
IN     91                 91
NY     90                 90
WI     88                 88
KY     87                 87
NV     85                 85
SC     80                 80
MD     77                 77
OR     76                 76
AR     73                 73
MI     71                 71
MS     61                 61
MN     60                 60
NJ     60                 60
UT     58                 58
KS     49                 49
WV     46                 46
ID     37                 37
AK     36                 36
MA     33                 33
IA     31                 31
HI     29                 29
MT     29                 29
NE     24                 24
ME     21                 21
CT     20                 20
SD     14                 14
DC     13                 13
WY     13                 13
NH     12                 12
ND     11                 11
DE     10                 10
VT     8                   8
RI     4                   4
Name: shooting count, dtype: int64

Exploratory Data Analysis (EDA)

data = dict(type='choropleth',
            locations = shooting['state'],
            locationmode = 'USA-states',
            colorscale = 'Reds',
            
            z = shooting['shooting count'],
            colorbar = {'title':"Cases Count"}
            )

layout = dict(title = 'Shooting Cases Per State in USA',
              geo = dict(scope='usa')
             )

choromap = go.Figure(data = [data],layout = layout)
iplot(choromap)

What do we see from our map?

• California has the highest number of shootings incidents and Rhode Island has the least with only 4 cases
• The 3 most populated states (California, Texas and Florida) have the most incidents.

plt.figure(figsize = (13,8))
sns.set_style('whitegrid')
shooting['age'].hist(bins=17, range=[10, 90], color = 'salmon')
plt.xlabel('age')
plt.ylabel('count')
plt.title('Shooting Case per Age Distribution')

Text(0.5, 1.0, 'Shooting Case per Age Distribution')

What does histogram tell us?

• The majority of shooting vistims are from 20 to 40 years old
• There are a few cases for senior victims and below 20

plt.figure(figsize = (13,8))

sns.countplot(x='gender',data= shooting,palette='coolwarm')
plt.xlabel('Gender')
plt.ylabel('Count')
plt.title('Shooting Case per Gender Distribution')

Text(0.5, 1.0, 'Shooting Case per Gender Distribution')

When distributed by gender, we can see Females are much less likely to be shot and killed by police then males.

words = ['armed']
word2 = pd.read_csv('shootings.csv',usecols=words)
wordcloud = WordCloud(background_color="white",max_font_size=100).generate(' '.join(word2['armed']))

plt.figure(figsize=(12,12))
plt.imshow(wordcloud)
plt.axis("off")
plt.show()

The majority of weapon that victims used in our dataset were guns, knives, toy weapons, and vehicles. Surprisingly, there were also a large number of unarmed and unknown weapons reported in our dataset.

Feature Engineering

shooting

	id	name	date	manner_of_death	armed	age	gender	race	city	state	signs_of_mental_illness	threat_level	flee	body_camera	arms_category	shooting count
0	3	Tim Elliot	2015-01-02	shot	gun	53.0	M	Asian	Shelton	WA	True	attack	Not fleeing	False	Guns	126
1	4	Lewis Lee Lembke	2015-01-02	shot	gun	47.0	M	White	Aloha	OR	False	attack	Not fleeing	False	Guns	76
2	5	John Paul Quintero	2015-01-03	shot and Tasered	unarmed	23.0	M	Hispanic	Wichita	KS	False	other	Not fleeing	False	Unarmed	49
3	8	Matthew Hoffman	2015-01-04	shot	toy weapon	32.0	M	White	San Francisco	CA	True	attack	Not fleeing	False	Other unusual objects	701
4	9	Michael Rodriguez	2015-01-04	shot	nail gun	39.0	M	Hispanic	Evans	CO	False	attack	Not fleeing	False	Piercing objects	168
...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...
4890	5916	Rayshard Brooks	2020-06-12	shot	Taser	27.0	M	Black	Atlanta	GA	False	attack	Foot	True	Electrical devices	161
4891	5925	Caine Van Pelt	2020-06-12	shot	gun	23.0	M	Black	Crown Point	IN	False	attack	Car	False	Guns	91
4892	5918	Hannah Fizer	2020-06-13	shot	unarmed	25.0	F	White	Sedalia	MO	False	other	Not fleeing	False	Unarmed	124
4893	5921	William Slyter	2020-06-13	shot	gun	22.0	M	White	Kansas City	MO	False	other	Other	False	Guns	124
4894	5924	Nicholas Hirsh	2020-06-15	shot	gun	31.0	M	White	Lawrence	KS	False	attack	Car	False	Guns	49

4895 rows × 16 columns

shooting['armed'].value_counts().sort_values(ascending=False).head(20)

gun                2755
knife               708
unknown             418
unarmed             348
toy weapon          171
vehicle             120
machete              39
Taser                24
sword                22
ax                   21
baseball bat         16
gun and knife        15
hammer               14
metal pipe           12
screwdriver          12
sharp object         11
box cutter           11
hatchet              11
gun and vehicle      10
gun and car           9
Name: armed, dtype: int64

top_10_labels = [y for y in shooting.armed.value_counts().sort_values(ascending=False).head(10).index]
top_10_labels

['gun',
 'knife',
 'unknown',
 'unarmed',
 'toy weapon',
 'vehicle',
 'machete',
 'Taser',
 'sword',
 'ax']

def one_hot_encoding_top_x(shooting, variable, top_x_labels):
    # function to create the dummy variables for the most frequent labels
    # we can vary the number of most frequent labels that we encode
    
    for label in top_x_labels:
        shooting[variable+'_'+label] = np.where(shooting[variable]==label, 1, 0)

# read the data again
shooting = pd.read_csv('shootings.csv')

# encode Xarmed into the 10 most frequent categories
one_hot_encoding_top_x(shooting, 'armed', top_10_labels)
shooting.head()

	id	name	date	manner_of_death	armed	age	gender	race	city	state	...	armed_gun	armed_knife	armed_unknown	armed_unarmed	armed_toy weapon	armed_vehicle	armed_machete	armed_Taser	armed_sword	armed_ax
0	3	Tim Elliot	2015-01-02	shot	gun	53.0	M	Asian	Shelton	WA	...	1	0	0	0	0	0	0	0	0	0
1	4	Lewis Lee Lembke	2015-01-02	shot	gun	47.0	M	White	Aloha	OR	...	1	0	0	0	0	0	0	0	0	0
2	5	John Paul Quintero	2015-01-03	shot and Tasered	unarmed	23.0	M	Hispanic	Wichita	KS	...	0	0	0	1	0	0	0	0	0	0
3	8	Matthew Hoffman	2015-01-04	shot	toy weapon	32.0	M	White	San Francisco	CA	...	0	0	0	0	1	0	0	0	0	0
4	9	Michael Rodriguez	2015-01-04	shot	nail gun	39.0	M	Hispanic	Evans	CO	...	0	0	0	0	0	0	0	0	0	0

5 rows × 25 columns

shooting['arms_category'].value_counts().sort_values(ascending=False).head(20)

Guns                     2764
Sharp objects             818
Unknown                   418
Unarmed                   348
Other unusual objects     192
Blunt instruments         122
Vehicles                  121
Multiple                   54
Piercing objects           29
Electrical devices         24
Explosives                  4
Hand tools                  1
Name: arms_category, dtype: int64

top_10_labels = [y for y in shooting.arms_category.value_counts().sort_values(ascending=False).head(10).index]
top_10_labels

['Guns',
 'Sharp objects',
 'Unknown',
 'Unarmed',
 'Other unusual objects',
 'Blunt instruments',
 'Vehicles',
 'Multiple',
 'Piercing objects',
 'Electrical devices']

# read the data again
shooting = pd.read_csv('shootings.csv')

# encode arms_category into the 10 most frequent categories
one_hot_encoding_top_x(shooting, 'arms_category', top_10_labels)
shooting.head()

	id	name	date	manner_of_death	armed	age	gender	race	city	state	...	arms_category_Guns	arms_category_Sharp objects	arms_category_Unknown	arms_category_Unarmed	arms_category_Other unusual objects	arms_category_Blunt instruments	arms_category_Vehicles	arms_category_Multiple	arms_category_Piercing objects	arms_category_Electrical devices
0	3	Tim Elliot	2015-01-02	shot	gun	53.0	M	Asian	Shelton	WA	...	1	0	0	0	0	0	0	0	0	0
1	4	Lewis Lee Lembke	2015-01-02	shot	gun	47.0	M	White	Aloha	OR	...	1	0	0	0	0	0	0	0	0	0
2	5	John Paul Quintero	2015-01-03	shot and Tasered	unarmed	23.0	M	Hispanic	Wichita	KS	...	0	0	0	1	0	0	0	0	0	0
3	8	Matthew Hoffman	2015-01-04	shot	toy weapon	32.0	M	White	San Francisco	CA	...	0	0	0	0	1	0	0	0	0	0
4	9	Michael Rodriguez	2015-01-04	shot	nail gun	39.0	M	Hispanic	Evans	CO	...	0	0	0	0	0	0	0	0	1	0

5 rows × 25 columns

shooting = shooting.drop('id', axis = 1)
shooting

	name	date	manner_of_death	armed	age	gender	race	city	state	signs_of_mental_illness	...	arms_category_Guns	arms_category_Sharp objects	arms_category_Unknown	arms_category_Unarmed	arms_category_Other unusual objects	arms_category_Blunt instruments	arms_category_Vehicles	arms_category_Multiple	arms_category_Piercing objects	arms_category_Electrical devices
0	Tim Elliot	2015-01-02	shot	gun	53.0	M	Asian	Shelton	WA	True	...	1	0	0	0	0	0	0	0	0	0
1	Lewis Lee Lembke	2015-01-02	shot	gun	47.0	M	White	Aloha	OR	False	...	1	0	0	0	0	0	0	0	0	0
2	John Paul Quintero	2015-01-03	shot and Tasered	unarmed	23.0	M	Hispanic	Wichita	KS	False	...	0	0	0	1	0	0	0	0	0	0
3	Matthew Hoffman	2015-01-04	shot	toy weapon	32.0	M	White	San Francisco	CA	True	...	0	0	0	0	1	0	0	0	0	0
4	Michael Rodriguez	2015-01-04	shot	nail gun	39.0	M	Hispanic	Evans	CO	False	...	0	0	0	0	0	0	0	0	1	0
...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...
4890	Rayshard Brooks	2020-06-12	shot	Taser	27.0	M	Black	Atlanta	GA	False	...	0	0	0	0	0	0	0	0	0	1
4891	Caine Van Pelt	2020-06-12	shot	gun	23.0	M	Black	Crown Point	IN	False	...	1	0	0	0	0	0	0	0	0	0
4892	Hannah Fizer	2020-06-13	shot	unarmed	25.0	F	White	Sedalia	MO	False	...	0	0	0	1	0	0	0	0	0	0
4893	William Slyter	2020-06-13	shot	gun	22.0	M	White	Kansas City	MO	False	...	1	0	0	0	0	0	0	0	0	0
4894	Nicholas Hirsh	2020-06-15	shot	gun	31.0	M	White	Lawrence	KS	False	...	1	0	0	0	0	0	0	0	0	0

4895 rows × 24 columns

shooting['gender'] = shooting['gender'].map({'M':0,'F':1})

shooting

	name	date	manner_of_death	armed	age	gender	race	city	state	signs_of_mental_illness	...	arms_category_Guns	arms_category_Sharp objects	arms_category_Unknown	arms_category_Unarmed	arms_category_Other unusual objects	arms_category_Blunt instruments	arms_category_Vehicles	arms_category_Multiple	arms_category_Piercing objects	arms_category_Electrical devices
0	Tim Elliot	2015-01-02	shot	gun	53.0	0	Asian	Shelton	WA	True	...	1	0	0	0	0	0	0	0	0	0
1	Lewis Lee Lembke	2015-01-02	shot	gun	47.0	0	White	Aloha	OR	False	...	1	0	0	0	0	0	0	0	0	0
2	John Paul Quintero	2015-01-03	shot and Tasered	unarmed	23.0	0	Hispanic	Wichita	KS	False	...	0	0	0	1	0	0	0	0	0	0
3	Matthew Hoffman	2015-01-04	shot	toy weapon	32.0	0	White	San Francisco	CA	True	...	0	0	0	0	1	0	0	0	0	0
4	Michael Rodriguez	2015-01-04	shot	nail gun	39.0	0	Hispanic	Evans	CO	False	...	0	0	0	0	0	0	0	0	1	0
...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...
4890	Rayshard Brooks	2020-06-12	shot	Taser	27.0	0	Black	Atlanta	GA	False	...	0	0	0	0	0	0	0	0	0	1
4891	Caine Van Pelt	2020-06-12	shot	gun	23.0	0	Black	Crown Point	IN	False	...	1	0	0	0	0	0	0	0	0	0
4892	Hannah Fizer	2020-06-13	shot	unarmed	25.0	1	White	Sedalia	MO	False	...	0	0	0	1	0	0	0	0	0	0
4893	William Slyter	2020-06-13	shot	gun	22.0	0	White	Kansas City	MO	False	...	1	0	0	0	0	0	0	0	0	0
4894	Nicholas Hirsh	2020-06-15	shot	gun	31.0	0	White	Lawrence	KS	False	...	1	0	0	0	0	0	0	0	0	0

4895 rows × 24 columns

shooting.groupby('race').size()

race
Asian         93
Black       1298
Hispanic     902
Native        78
Other         48
White       2476
dtype: int64

race_dummies = pd.get_dummies(shooting['race'], drop_first=False)
shooting = pd.concat([shooting,race_dummies],axis=1)

shooting

	name	date	manner_of_death	armed	age	gender	race	city	state	signs_of_mental_illness	...	arms_category_Vehicles	arms_category_Multiple	arms_category_Piercing objects	arms_category_Electrical devices	Asian	Black	Hispanic	Native	Other	White
0	Tim Elliot	2015-01-02	shot	gun	53.0	0	Asian	Shelton	WA	True	...	0	0	0	0	1	0	0	0	0	0
1	Lewis Lee Lembke	2015-01-02	shot	gun	47.0	0	White	Aloha	OR	False	...	0	0	0	0	0	0	0	0	0	1
2	John Paul Quintero	2015-01-03	shot and Tasered	unarmed	23.0	0	Hispanic	Wichita	KS	False	...	0	0	0	0	0	0	1	0	0	0
3	Matthew Hoffman	2015-01-04	shot	toy weapon	32.0	0	White	San Francisco	CA	True	...	0	0	0	0	0	0	0	0	0	1
4	Michael Rodriguez	2015-01-04	shot	nail gun	39.0	0	Hispanic	Evans	CO	False	...	0	0	1	0	0	0	1	0	0	0
...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...
4890	Rayshard Brooks	2020-06-12	shot	Taser	27.0	0	Black	Atlanta	GA	False	...	0	0	0	1	0	1	0	0	0	0
4891	Caine Van Pelt	2020-06-12	shot	gun	23.0	0	Black	Crown Point	IN	False	...	0	0	0	0	0	1	0	0	0	0
4892	Hannah Fizer	2020-06-13	shot	unarmed	25.0	1	White	Sedalia	MO	False	...	0	0	0	0	0	0	0	0	0	1
4893	William Slyter	2020-06-13	shot	gun	22.0	0	White	Kansas City	MO	False	...	0	0	0	0	0	0	0	0	0	1
4894	Nicholas Hirsh	2020-06-15	shot	gun	31.0	0	White	Lawrence	KS	False	...	0	0	0	0	0	0	0	0	0	1

4895 rows × 30 columns

shooting = shooting.drop('race', axis = 1)

shooting

	name	date	manner_of_death	armed	age	gender	city	state	signs_of_mental_illness	threat_level	...	arms_category_Vehicles	arms_category_Multiple	arms_category_Piercing objects	arms_category_Electrical devices	Asian	Black	Hispanic	Native	Other	White
0	Tim Elliot	2015-01-02	shot	gun	53.0	0	Shelton	WA	True	attack	...	0	0	0	0	1	0	0	0	0	0
1	Lewis Lee Lembke	2015-01-02	shot	gun	47.0	0	Aloha	OR	False	attack	...	0	0	0	0	0	0	0	0	0	1
2	John Paul Quintero	2015-01-03	shot and Tasered	unarmed	23.0	0	Wichita	KS	False	other	...	0	0	0	0	0	0	1	0	0	0
3	Matthew Hoffman	2015-01-04	shot	toy weapon	32.0	0	San Francisco	CA	True	attack	...	0	0	0	0	0	0	0	0	0	1
4	Michael Rodriguez	2015-01-04	shot	nail gun	39.0	0	Evans	CO	False	attack	...	0	0	1	0	0	0	1	0	0	0
...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...
4890	Rayshard Brooks	2020-06-12	shot	Taser	27.0	0	Atlanta	GA	False	attack	...	0	0	0	1	0	1	0	0	0	0
4891	Caine Van Pelt	2020-06-12	shot	gun	23.0	0	Crown Point	IN	False	attack	...	0	0	0	0	0	1	0	0	0	0
4892	Hannah Fizer	2020-06-13	shot	unarmed	25.0	1	Sedalia	MO	False	other	...	0	0	0	0	0	0	0	0	0	1
4893	William Slyter	2020-06-13	shot	gun	22.0	0	Kansas City	MO	False	other	...	0	0	0	0	0	0	0	0	0	1
4894	Nicholas Hirsh	2020-06-15	shot	gun	31.0	0	Lawrence	KS	False	attack	...	0	0	0	0	0	0	0	0	0	1

4895 rows × 29 columns

# 0 as shot and 1 as shot and Tasered
shooting['manner_of_death'] = shooting['manner_of_death'].map({'shot':0, 'shot and Tasered':1})

shooting

	name	date	manner_of_death	armed	age	gender	city	state	signs_of_mental_illness	threat_level	...	arms_category_Vehicles	arms_category_Multiple	arms_category_Piercing objects	arms_category_Electrical devices	Asian	Black	Hispanic	Native	Other	White
0	Tim Elliot	2015-01-02	0	gun	53.0	0	Shelton	WA	True	attack	...	0	0	0	0	1	0	0	0	0	0
1	Lewis Lee Lembke	2015-01-02	0	gun	47.0	0	Aloha	OR	False	attack	...	0	0	0	0	0	0	0	0	0	1
2	John Paul Quintero	2015-01-03	1	unarmed	23.0	0	Wichita	KS	False	other	...	0	0	0	0	0	0	1	0	0	0
3	Matthew Hoffman	2015-01-04	0	toy weapon	32.0	0	San Francisco	CA	True	attack	...	0	0	0	0	0	0	0	0	0	1
4	Michael Rodriguez	2015-01-04	0	nail gun	39.0	0	Evans	CO	False	attack	...	0	0	1	0	0	0	1	0	0	0
...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...
4890	Rayshard Brooks	2020-06-12	0	Taser	27.0	0	Atlanta	GA	False	attack	...	0	0	0	1	0	1	0	0	0	0
4891	Caine Van Pelt	2020-06-12	0	gun	23.0	0	Crown Point	IN	False	attack	...	0	0	0	0	0	1	0	0	0	0
4892	Hannah Fizer	2020-06-13	0	unarmed	25.0	1	Sedalia	MO	False	other	...	0	0	0	0	0	0	0	0	0	1
4893	William Slyter	2020-06-13	0	gun	22.0	0	Kansas City	MO	False	other	...	0	0	0	0	0	0	0	0	0	1
4894	Nicholas Hirsh	2020-06-15	0	gun	31.0	0	Lawrence	KS	False	attack	...	0	0	0	0	0	0	0	0	0	1

4895 rows × 29 columns

shooting

	name	date	manner_of_death	armed	age	gender	city	state	signs_of_mental_illness	threat_level	...	arms_category_Vehicles	arms_category_Multiple	arms_category_Piercing objects	arms_category_Electrical devices	Asian	Black	Hispanic	Native	Other	White
0	Tim Elliot	2015-01-02	0	gun	53.0	0	Shelton	WA	True	attack	...	0	0	0	0	1	0	0	0	0	0
1	Lewis Lee Lembke	2015-01-02	0	gun	47.0	0	Aloha	OR	False	attack	...	0	0	0	0	0	0	0	0	0	1
2	John Paul Quintero	2015-01-03	1	unarmed	23.0	0	Wichita	KS	False	other	...	0	0	0	0	0	0	1	0	0	0
3	Matthew Hoffman	2015-01-04	0	toy weapon	32.0	0	San Francisco	CA	True	attack	...	0	0	0	0	0	0	0	0	0	1
4	Michael Rodriguez	2015-01-04	0	nail gun	39.0	0	Evans	CO	False	attack	...	0	0	1	0	0	0	1	0	0	0
...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...
4890	Rayshard Brooks	2020-06-12	0	Taser	27.0	0	Atlanta	GA	False	attack	...	0	0	0	1	0	1	0	0	0	0
4891	Caine Van Pelt	2020-06-12	0	gun	23.0	0	Crown Point	IN	False	attack	...	0	0	0	0	0	1	0	0	0	0
4892	Hannah Fizer	2020-06-13	0	unarmed	25.0	1	Sedalia	MO	False	other	...	0	0	0	0	0	0	0	0	0	1
4893	William Slyter	2020-06-13	0	gun	22.0	0	Kansas City	MO	False	other	...	0	0	0	0	0	0	0	0	0	1
4894	Nicholas Hirsh	2020-06-15	0	gun	31.0	0	Lawrence	KS	False	attack	...	0	0	0	0	0	0	0	0	0	1

4895 rows × 29 columns

shooting['signs_of_mental_illness'] = shooting['signs_of_mental_illness'].map({False:0, True:1}).astype(int)

shooting['threat_level'] = shooting['threat_level'].map({'attack':1, 'other':0, 'undetermined':2})

shooting

	name	date	manner_of_death	armed	age	gender	city	state	signs_of_mental_illness	threat_level	...	arms_category_Vehicles	arms_category_Multiple	arms_category_Piercing objects	arms_category_Electrical devices	Asian	Black	Hispanic	Native	Other	White
0	Tim Elliot	2015-01-02	0	gun	53.0	0	Shelton	WA	1	1	...	0	0	0	0	1	0	0	0	0	0
1	Lewis Lee Lembke	2015-01-02	0	gun	47.0	0	Aloha	OR	0	1	...	0	0	0	0	0	0	0	0	0	1
2	John Paul Quintero	2015-01-03	1	unarmed	23.0	0	Wichita	KS	0	0	...	0	0	0	0	0	0	1	0	0	0
3	Matthew Hoffman	2015-01-04	0	toy weapon	32.0	0	San Francisco	CA	1	1	...	0	0	0	0	0	0	0	0	0	1
4	Michael Rodriguez	2015-01-04	0	nail gun	39.0	0	Evans	CO	0	1	...	0	0	1	0	0	0	1	0	0	0
...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...
4890	Rayshard Brooks	2020-06-12	0	Taser	27.0	0	Atlanta	GA	0	1	...	0	0	0	1	0	1	0	0	0	0
4891	Caine Van Pelt	2020-06-12	0	gun	23.0	0	Crown Point	IN	0	1	...	0	0	0	0	0	1	0	0	0	0
4892	Hannah Fizer	2020-06-13	0	unarmed	25.0	1	Sedalia	MO	0	0	...	0	0	0	0	0	0	0	0	0	1
4893	William Slyter	2020-06-13	0	gun	22.0	0	Kansas City	MO	0	0	...	0	0	0	0	0	0	0	0	0	1
4894	Nicholas Hirsh	2020-06-15	0	gun	31.0	0	Lawrence	KS	0	1	...	0	0	0	0	0	0	0	0	0	1

4895 rows × 29 columns

shooting.groupby('flee').size()

flee
Car             820
Foot            642
Not fleeing    3073
Other           360
dtype: int64

flee_dummies = pd.get_dummies(shooting['flee'], drop_first=False)
shooting = pd.concat([shooting,flee_dummies],axis=1)

shooting = shooting.drop('flee', axis =1)

shooting.iloc[:,0:18]

	name	date	manner_of_death	armed	age	gender	city	state	signs_of_mental_illness	threat_level	body_camera	arms_category	arms_category_Guns	arms_category_Sharp objects	arms_category_Unknown	arms_category_Unarmed	arms_category_Other unusual objects	arms_category_Blunt instruments
0	Tim Elliot	2015-01-02	0	gun	53.0	0	Shelton	WA	1	1	False	Guns	1	0	0	0	0	0
1	Lewis Lee Lembke	2015-01-02	0	gun	47.0	0	Aloha	OR	0	1	False	Guns	1	0	0	0	0	0
2	John Paul Quintero	2015-01-03	1	unarmed	23.0	0	Wichita	KS	0	0	False	Unarmed	0	0	0	1	0	0
3	Matthew Hoffman	2015-01-04	0	toy weapon	32.0	0	San Francisco	CA	1	1	False	Other unusual objects	0	0	0	0	1	0
4	Michael Rodriguez	2015-01-04	0	nail gun	39.0	0	Evans	CO	0	1	False	Piercing objects	0	0	0	0	0	0
...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...
4890	Rayshard Brooks	2020-06-12	0	Taser	27.0	0	Atlanta	GA	0	1	True	Electrical devices	0	0	0	0	0	0
4891	Caine Van Pelt	2020-06-12	0	gun	23.0	0	Crown Point	IN	0	1	False	Guns	1	0	0	0	0	0
4892	Hannah Fizer	2020-06-13	0	unarmed	25.0	1	Sedalia	MO	0	0	False	Unarmed	0	0	0	1	0	0
4893	William Slyter	2020-06-13	0	gun	22.0	0	Kansas City	MO	0	0	False	Guns	1	0	0	0	0	0
4894	Nicholas Hirsh	2020-06-15	0	gun	31.0	0	Lawrence	KS	0	1	False	Guns	1	0	0	0	0	0

4895 rows × 18 columns

shooting['body_camera'] = shooting['body_camera'].map({True:1, False:0}).astype(int)

shooting['body_camera']

0       0
1       0
2       0
3       0
4       0
       ..
4890    1
4891    0
4892    0
4893    0
4894    0
Name: body_camera, Length: 4895, dtype: int32

arms_cat_dummies = pd.get_dummies(shooting['arms_category'], drop_first=False)
shooting = pd.concat([shooting,arms_cat_dummies],axis=1)

shooting = shooting.drop('arms_category', axis = 1)

shooting

	name	date	manner_of_death	armed	age	gender	city	state	signs_of_mental_illness	threat_level	...	Explosives	Guns	Hand tools	Multiple	Other unusual objects	Piercing objects	Sharp objects	Unarmed	Unknown	Vehicles
0	Tim Elliot	2015-01-02	0	gun	53.0	0	Shelton	WA	1	1	...	0	1	0	0	0	0	0	0	0	0
1	Lewis Lee Lembke	2015-01-02	0	gun	47.0	0	Aloha	OR	0	1	...	0	1	0	0	0	0	0	0	0	0
2	John Paul Quintero	2015-01-03	1	unarmed	23.0	0	Wichita	KS	0	0	...	0	0	0	0	0	0	0	1	0	0
3	Matthew Hoffman	2015-01-04	0	toy weapon	32.0	0	San Francisco	CA	1	1	...	0	0	0	0	1	0	0	0	0	0
4	Michael Rodriguez	2015-01-04	0	nail gun	39.0	0	Evans	CO	0	1	...	0	0	0	0	0	1	0	0	0	0
...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...
4890	Rayshard Brooks	2020-06-12	0	Taser	27.0	0	Atlanta	GA	0	1	...	0	0	0	0	0	0	0	0	0	0
4891	Caine Van Pelt	2020-06-12	0	gun	23.0	0	Crown Point	IN	0	1	...	0	1	0	0	0	0	0	0	0	0
4892	Hannah Fizer	2020-06-13	0	unarmed	25.0	1	Sedalia	MO	0	0	...	0	0	0	0	0	0	0	1	0	0
4893	William Slyter	2020-06-13	0	gun	22.0	0	Kansas City	MO	0	0	...	0	1	0	0	0	0	0	0	0	0
4894	Nicholas Hirsh	2020-06-15	0	gun	31.0	0	Lawrence	KS	0	1	...	0	1	0	0	0	0	0	0	0	0

4895 rows × 43 columns

shooting = shooting.drop(['state', 'city','name'], axis = 1)

shooting

	date	manner_of_death	armed	age	gender	signs_of_mental_illness	threat_level	body_camera	arms_category_Guns	arms_category_Sharp objects	...	Explosives	Guns	Hand tools	Multiple	Other unusual objects	Piercing objects	Sharp objects	Unarmed	Unknown	Vehicles
0	2015-01-02	0	gun	53.0	0	1	1	0	1	0	...	0	1	0	0	0	0	0	0	0	0
1	2015-01-02	0	gun	47.0	0	0	1	0	1	0	...	0	1	0	0	0	0	0	0	0	0
2	2015-01-03	1	unarmed	23.0	0	0	0	0	0	0	...	0	0	0	0	0	0	0	1	0	0
3	2015-01-04	0	toy weapon	32.0	0	1	1	0	0	0	...	0	0	0	0	1	0	0	0	0	0
4	2015-01-04	0	nail gun	39.0	0	0	1	0	0	0	...	0	0	0	0	0	1	0	0	0	0
...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...
4890	2020-06-12	0	Taser	27.0	0	0	1	1	0	0	...	0	0	0	0	0	0	0	0	0	0
4891	2020-06-12	0	gun	23.0	0	0	1	0	1	0	...	0	1	0	0	0	0	0	0	0	0
4892	2020-06-13	0	unarmed	25.0	1	0	0	0	0	0	...	0	0	0	0	0	0	0	1	0	0
4893	2020-06-13	0	gun	22.0	0	0	0	0	1	0	...	0	1	0	0	0	0	0	0	0	0
4894	2020-06-15	0	gun	31.0	0	0	1	0	1	0	...	0	1	0	0	0	0	0	0	0	0

4895 rows × 40 columns

shooting['date'] = pd.to_datetime(shooting['date'])
shooting['day'] = shooting['date'].apply(lambda date:date.day)
shooting['month'] = shooting['date'].apply(lambda date:date.month)
shooting['year'] = shooting['date'].apply(lambda date:date.year)

<ipython-input-51-b3b118feae7d>:1: SettingWithCopyWarning:


A value is trying to be set on a copy of a slice from a DataFrame.
Try using .loc[row_indexer,col_indexer] = value instead

See the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy

<ipython-input-51-b3b118feae7d>:2: SettingWithCopyWarning:


A value is trying to be set on a copy of a slice from a DataFrame.
Try using .loc[row_indexer,col_indexer] = value instead

See the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy

<ipython-input-51-b3b118feae7d>:3: SettingWithCopyWarning:


A value is trying to be set on a copy of a slice from a DataFrame.
Try using .loc[row_indexer,col_indexer] = value instead

See the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy

shooting

	date	manner_of_death	armed	age	gender	signs_of_mental_illness	threat_level	body_camera	arms_category_Guns	arms_category_Sharp objects	...	Multiple	Other unusual objects	Piercing objects	Sharp objects	Unarmed	Unknown	Vehicles	day	month	year
0	2015-01-02	0	gun	53.0	0	1	1	0	1	0	...	0	0	0	0	0	0	0	2	1	2015
1	2015-01-02	0	gun	47.0	0	0	1	0	1	0	...	0	0	0	0	0	0	0	2	1	2015
2	2015-01-03	1	unarmed	23.0	0	0	0	0	0	0	...	0	0	0	0	1	0	0	3	1	2015
3	2015-01-04	0	toy weapon	32.0	0	1	1	0	0	0	...	0	1	0	0	0	0	0	4	1	2015
4	2015-01-04	0	nail gun	39.0	0	0	1	0	0	0	...	0	0	1	0	0	0	0	4	1	2015
...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...
4890	2020-06-12	0	Taser	27.0	0	0	1	1	0	0	...	0	0	0	0	0	0	0	12	6	2020
4891	2020-06-12	0	gun	23.0	0	0	1	0	1	0	...	0	0	0	0	0	0	0	12	6	2020
4892	2020-06-13	0	unarmed	25.0	1	0	0	0	0	0	...	0	0	0	0	1	0	0	13	6	2020
4893	2020-06-13	0	gun	22.0	0	0	0	0	1	0	...	0	0	0	0	0	0	0	13	6	2020
4894	2020-06-15	0	gun	31.0	0	0	1	0	1	0	...	0	0	0	0	0	0	0	15	6	2020

4895 rows × 43 columns

shooting = shooting.drop('date',axis =1)

shooting

	manner_of_death	armed	age	gender	signs_of_mental_illness	threat_level	body_camera	arms_category_Guns	arms_category_Sharp objects	arms_category_Unknown	...	Multiple	Other unusual objects	Piercing objects	Sharp objects	Unarmed	Unknown	Vehicles	day	month	year
0	0	gun	53.0	0	1	1	0	1	0	0	...	0	0	0	0	0	0	0	2	1	2015
1	0	gun	47.0	0	0	1	0	1	0	0	...	0	0	0	0	0	0	0	2	1	2015
2	1	unarmed	23.0	0	0	0	0	0	0	0	...	0	0	0	0	1	0	0	3	1	2015
3	0	toy weapon	32.0	0	1	1	0	0	0	0	...	0	1	0	0	0	0	0	4	1	2015
4	0	nail gun	39.0	0	0	1	0	0	0	0	...	0	0	1	0	0	0	0	4	1	2015
...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...
4890	0	Taser	27.0	0	0	1	1	0	0	0	...	0	0	0	0	0	0	0	12	6	2020
4891	0	gun	23.0	0	0	1	0	1	0	0	...	0	0	0	0	0	0	0	12	6	2020
4892	0	unarmed	25.0	1	0	0	0	0	0	0	...	0	0	0	0	1	0	0	13	6	2020
4893	0	gun	22.0	0	0	0	0	1	0	0	...	0	0	0	0	0	0	0	13	6	2020
4894	0	gun	31.0	0	0	1	0	1	0	0	...	0	0	0	0	0	0	0	15	6	2020

4895 rows × 42 columns

shooting = shooting.drop('armed',axis = 1)

shooting.columns

Index(['manner_of_death', 'age', 'gender', 'signs_of_mental_illness',
       'threat_level', 'body_camera', 'arms_category_Guns',
       'arms_category_Sharp objects', 'arms_category_Unknown',
       'arms_category_Unarmed', 'arms_category_Other unusual objects',
       'arms_category_Blunt instruments', 'arms_category_Vehicles',
       'arms_category_Multiple', 'arms_category_Piercing objects',
       'arms_category_Electrical devices', 'Asian', 'Black', 'Hispanic',
       'Native', 'Other', 'White', 'Car', 'Foot', 'Not fleeing', 'Other',
       'Blunt instruments', 'Electrical devices', 'Explosives', 'Guns',
       'Hand tools', 'Multiple', 'Other unusual objects', 'Piercing objects',
       'Sharp objects', 'Unarmed', 'Unknown', 'Vehicles', 'day', 'month',
       'year'],
      dtype='object')

shooting.info()

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 4895 entries, 0 to 4894
Data columns (total 41 columns):
 #   Column                               Non-Null Count  Dtype  
---  ------                               --------------  -----  
 0   manner_of_death                      4895 non-null   int64  
 1   age                                  4895 non-null   float64
 2   gender                               4895 non-null   int64  
 3   signs_of_mental_illness              4895 non-null   int32  
 4   threat_level                         4895 non-null   int64  
 5   body_camera                          4895 non-null   int32  
 6   arms_category_Guns                   4895 non-null   int32  
 7   arms_category_Sharp objects          4895 non-null   int32  
 8   arms_category_Unknown                4895 non-null   int32  
 9   arms_category_Unarmed                4895 non-null   int32  
 10  arms_category_Other unusual objects  4895 non-null   int32  
 11  arms_category_Blunt instruments      4895 non-null   int32  
 12  arms_category_Vehicles               4895 non-null   int32  
 13  arms_category_Multiple               4895 non-null   int32  
 14  arms_category_Piercing objects       4895 non-null   int32  
 15  arms_category_Electrical devices     4895 non-null   int32  
 16  Asian                                4895 non-null   uint8  
 17  Black                                4895 non-null   uint8  
 18  Hispanic                             4895 non-null   uint8  
 19  Native                               4895 non-null   uint8  
 20  Other                                4895 non-null   uint8  
 21  White                                4895 non-null   uint8  
 22  Car                                  4895 non-null   uint8  
 23  Foot                                 4895 non-null   uint8  
 24  Not fleeing                          4895 non-null   uint8  
 25  Other                                4895 non-null   uint8  
 26  Blunt instruments                    4895 non-null   uint8  
 27  Electrical devices                   4895 non-null   uint8  
 28  Explosives                           4895 non-null   uint8  
 29  Guns                                 4895 non-null   uint8  
 30  Hand tools                           4895 non-null   uint8  
 31  Multiple                             4895 non-null   uint8  
 32  Other unusual objects                4895 non-null   uint8  
 33  Piercing objects                     4895 non-null   uint8  
 34  Sharp objects                        4895 non-null   uint8  
 35  Unarmed                              4895 non-null   uint8  
 36  Unknown                              4895 non-null   uint8  
 37  Vehicles                             4895 non-null   uint8  
 38  day                                  4895 non-null   int64  
 39  month                                4895 non-null   int64  
 40  year                                 4895 non-null   int64  
dtypes: float64(1), int32(12), int64(6), uint8(22)
memory usage: 602.4 KB

shooting

	manner_of_death	age	gender	signs_of_mental_illness	threat_level	body_camera	arms_category_Guns	arms_category_Sharp objects	arms_category_Unknown	arms_category_Unarmed	...	Multiple	Other unusual objects	Piercing objects	Sharp objects	Unarmed	Unknown	Vehicles	day	month	year
0	0	53.0	0	1	1	0	1	0	0	0	...	0	0	0	0	0	0	0	2	1	2015
1	0	47.0	0	0	1	0	1	0	0	0	...	0	0	0	0	0	0	0	2	1	2015
2	1	23.0	0	0	0	0	0	0	0	1	...	0	0	0	0	1	0	0	3	1	2015
3	0	32.0	0	1	1	0	0	0	0	0	...	0	1	0	0	0	0	0	4	1	2015
4	0	39.0	0	0	1	0	0	0	0	0	...	0	0	1	0	0	0	0	4	1	2015
...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...
4890	0	27.0	0	0	1	1	0	0	0	0	...	0	0	0	0	0	0	0	12	6	2020
4891	0	23.0	0	0	1	0	1	0	0	0	...	0	0	0	0	0	0	0	12	6	2020
4892	0	25.0	1	0	0	0	0	0	0	1	...	0	0	0	0	1	0	0	13	6	2020
4893	0	22.0	0	0	0	0	1	0	0	0	...	0	0	0	0	0	0	0	13	6	2020
4894	0	31.0	0	0	1	0	1	0	0	0	...	0	0	0	0	0	0	0	15	6	2020

4895 rows × 41 columns

PCA

std_scaler = StandardScaler()
std_scaler.fit(shooting)

StandardScaler()

pca = PCA(n_components=28)

scaled_data = std_scaler.transform(shooting)

pca.fit(scaled_data)

PCA(n_components=28)

scaled_data.shape

(4895, 41)

x_pca = pca.transform(scaled_data)

x_pca.shape

(4895, 28)

pca.components_

array([[ 1.58665094e-01, -4.17781576e-02,  3.84941665e-02, ...,
        -2.60396986e-04, -5.43632131e-03, -2.40555784e-02],
       [-7.23112520e-02, -1.77709848e-01,  1.25697512e-02, ...,
         3.36141947e-02,  1.15954326e-02,  4.43426268e-02],
       [ 7.20212199e-02, -1.46183080e-01, -6.94242563e-03, ...,
        -1.06707136e-02, -6.37247156e-02,  3.25601848e-02],
       ...,
       [-8.88904140e-02, -7.93972800e-02,  6.67531017e-02, ...,
        -1.13267946e-02, -9.76975112e-02,  2.78377205e-01],
       [-4.62211121e-02,  8.01240086e-01, -4.51234352e-03, ...,
         9.10206149e-02,  7.70129389e-02, -6.17455872e-02],
       [-2.88940371e-01,  1.86996106e-01,  5.14006269e-02, ...,
        -9.03868687e-02, -4.65529555e-01, -6.07363810e-01]])

variance = np.array(pca.explained_variance_ratio_)

list(map('{:.2f}'.format,variance))

['0.08',
 '0.07',
 '0.06',
 '0.05',
 '0.05',
 '0.05',
 '0.05',
 '0.05',
 '0.05',
 '0.05',
 '0.03',
 '0.03',
 '0.03',
 '0.03',
 '0.03',
 '0.03',
 '0.03',
 '0.02',
 '0.02',
 '0.02',
 '0.02',
 '0.02',
 '0.02',
 '0.02',
 '0.02',
 '0.02',
 '0.02',
 '0.02']

df_comp = pd.DataFrame(pca.components_,columns=shooting.columns)

plt.figure(figsize=(12,8))
sns.heatmap(df_comp,cmap='plasma',)

<AxesSubplot:>

As we see, the first 28 pca cover almost 100% variance of the data. we decided to drop the last 13 pca from 41 original one to avoid overfitting and increase model efficiency.

Machine Learning Models

Train-Test Split

shooting.keys()

Index(['manner_of_death', 'age', 'gender', 'signs_of_mental_illness',
       'threat_level', 'body_camera', 'arms_category_Guns',
       'arms_category_Sharp objects', 'arms_category_Unknown',
       'arms_category_Unarmed', 'arms_category_Other unusual objects',
       'arms_category_Blunt instruments', 'arms_category_Vehicles',
       'arms_category_Multiple', 'arms_category_Piercing objects',
       'arms_category_Electrical devices', 'Asian', 'Black', 'Hispanic',
       'Native', 'Other', 'White', 'Car', 'Foot', 'Not fleeing', 'Other',
       'Blunt instruments', 'Electrical devices', 'Explosives', 'Guns',
       'Hand tools', 'Multiple', 'Other unusual objects', 'Piercing objects',
       'Sharp objects', 'Unarmed', 'Unknown', 'Vehicles', 'day', 'month',
       'year'],
      dtype='object')

shooting.shape

(4895, 41)

X = shooting.drop('manner_of_death', axis =1)
y = shooting['manner_of_death']
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.3, random_state=0)

X.shape

(4895, 40)

y.shape

(4895,)

#fit pca into our train test
pca.fit(X_train)
X_train = pca.transform(X_train)
X_test = pca.transform(X_test)

logreg = LogisticRegression(random_state = 0, solver = 'liblinear',multi_class='auto') 
logreg_parameters = {'penalty':('l1', 'l2'), 'C':[0.01, 0.1, 1, 10, 100]}
logreg_grid = GridSearchCV(logreg, logreg_parameters,cv=10)
logreg_grid.fit(X_train, y_train)

GridSearchCV(cv=10,
             estimator=LogisticRegression(random_state=0, solver='liblinear'),
             param_grid={'C': [0.01, 0.1, 1, 10, 100], 'penalty': ('l1', 'l2')})

logreg_grid.best_params_

{'C': 0.01, 'penalty': 'l1'}

logreg_test = LogisticRegression(random_state = 0, solver = 'liblinear',multi_class='auto',C=0.01,penalty='l1')
logreg_test.fit(X_train, y_train)
print('Train score: {:.4f}'.format(logreg_test.score(X_train, y_train)))
print('Test score: {:.4f}'.format(logreg_test.score(X_test, y_test)))

Train score: 0.9521
Test score: 0.9428

from sklearn.neighbors import KNeighborsClassifier

knn = KNeighborsClassifier()
knn_parameters = {'n_neighbors':np.arange(1,20,1)}
knn_grid = GridSearchCV(knn, knn_parameters,cv=10)
knn_grid.fit(X_train, y_train)

GridSearchCV(cv=10, estimator=KNeighborsClassifier(),
             param_grid={'n_neighbors': array([ 1,  2,  3,  4,  5,  6,  7,  8,  9, 10, 11, 12, 13, 14, 15, 16, 17,
       18, 19])})

knn_grid.best_params_

{'n_neighbors': 2}

knn_test = KNeighborsClassifier(n_neighbors=2, leaf_size= 30, metric='minkowski', p=2, weights= 'uniform')
knn_test.fit(X_train, y_train)
print('Train score: {:.4f}'.format(knn_test.score(X_train, y_train)))
print('Test score: {:.4f}'.format(knn_test.score(X_test, y_test)))

Train score: 0.9545
Test score: 0.9415

knn_prediction = knn_test.predict(X_test)

print('Confusion Matrix \n', confusion_matrix(y_test, knn_prediction))
print(classification_report(y_test, knn_prediction))
print('Precision score: ', precision_score( y_test, knn_prediction, average = 'macro'))
print('Recall Score: ', recall_score( y_test, knn_prediction, average = 'macro'))

Confusion Matrix 
 [[1383    2]
 [  84    0]]
              precision    recall  f1-score   support

           0       0.94      1.00      0.97      1385
           1       0.00      0.00      0.00        84

    accuracy                           0.94      1469
   macro avg       0.47      0.50      0.48      1469
weighted avg       0.89      0.94      0.91      1469

Precision score:  0.47137014314928427
Recall Score:  0.49927797833935017

from sklearn.svm import LinearSVC

linr = LinearSVC(random_state = 0, dual=False)
linr_parameters = {'penalty':('l1', 'l2'), 'C':[0.001,0.01,0.1,1,10,100] }
linr_grid = GridSearchCV(linr, linr_parameters,cv=10,iid=True)
linr_grid.fit(X_train, y_train)

C:\Users\vochi\anaconda3\lib\site-packages\sklearn\model_selection\_search.py:847: FutureWarning:

The parameter 'iid' is deprecated in 0.22 and will be removed in 0.24.

GridSearchCV(cv=10, estimator=LinearSVC(dual=False, random_state=0), iid=True,
             param_grid={'C': [0.001, 0.01, 0.1, 1, 10, 100],
                         'penalty': ('l1', 'l2')})

linr_grid.best_params_

{'C': 0.001, 'penalty': 'l1'}

linr_test = LinearSVC(random_state = 0,dual=False,C= 0.001, penalty= 'l1')
linr_test.fit(X_train, y_train)
print('Train score: {:.4f}'.format(linr_test.score(X_train, y_train)))
print('Test score: {:.4f}'.format(linr_test.score(X_test, y_test)))

Train score: 0.9521
Test score: 0.9428

lnsvc_prediction = linr_test.predict(X_test)

print('Confusion Matrix \n', confusion_matrix(y_test, lnsvc_prediction))
print(classification_report(y_test, lnsvc_prediction))
print('Precision score: ', precision_score( y_test, lnsvc_prediction, average = 'macro'))
print('Recall Score: ', recall_score( y_test, lnsvc_prediction, average = 'macro'))

Confusion Matrix 
 [[1385    0]
 [  84    0]]
              precision    recall  f1-score   support

           0       0.94      1.00      0.97      1385
           1       0.00      0.00      0.00        84

    accuracy                           0.94      1469
   macro avg       0.47      0.50      0.49      1469
weighted avg       0.89      0.94      0.92      1469

Precision score:  0.4714091218515997
Recall Score:  0.5

C:\Users\vochi\anaconda3\lib\site-packages\sklearn\metrics\_classification.py:1221: UndefinedMetricWarning:

Precision and F-score are ill-defined and being set to 0.0 in labels with no predicted samples. Use `zero_division` parameter to control this behavior.

C:\Users\vochi\anaconda3\lib\site-packages\sklearn\metrics\_classification.py:1221: UndefinedMetricWarning:

Precision is ill-defined and being set to 0.0 in labels with no predicted samples. Use `zero_division` parameter to control this behavior.

from sklearn.svm import SVC

svm=SVC(kernel='linear',random_state=0)
#svm_parameters = {'gamma': [0.001, 0.01, 0.1, 1], 'C': [0.001, 0.01, 0.1, 1]}
#svm_grid = GridSearchCV(svm, svm_parameters,cv=3, n_jobs=4)
#svm_grid.fit(X_train, y_train)

svm_grid_para = {'C': 0.001,'gamma':0.001}
svm_grid_para
#C 0.001; Gamma 0.001

{'C': 0.001, 'gamma': 0.001}

from sklearn.svm import SVC
svm_test = SVC(kernel='linear',random_state=0,C= 0.001, gamma= 0.001)
svm_test.fit(X_train, y_train)
print('Train score: {:.4f}'.format(svm_test.score(X_train, y_train)))
print('Test score: {:.4f}'.format(svm_test.score(X_test, y_test)))

Train score: 0.9521
Test score: 0.9428

svm_prediction = svm_test.predict(X_test)

print('Confusion Matrix \n', confusion_matrix(y_test, svm_prediction))
print(classification_report(y_test, svm_prediction))
print('Precision score: ', precision_score( y_test, svm_prediction, average = 'macro'))
print('Recall Score: ', recall_score( y_test, svm_prediction, average = 'macro'))

Confusion Matrix 
 [[1385    0]
 [  84    0]]
              precision    recall  f1-score   support

           0       0.94      1.00      0.97      1385
           1       0.00      0.00      0.00        84

    accuracy                           0.94      1469
   macro avg       0.47      0.50      0.49      1469
weighted avg       0.89      0.94      0.92      1469

Precision score:  0.4714091218515997
Recall Score:  0.5

C:\Users\vochi\anaconda3\lib\site-packages\sklearn\metrics\_classification.py:1221: UndefinedMetricWarning:

Precision and F-score are ill-defined and being set to 0.0 in labels with no predicted samples. Use `zero_division` parameter to control this behavior.

C:\Users\vochi\anaconda3\lib\site-packages\sklearn\metrics\_classification.py:1221: UndefinedMetricWarning:

Precision is ill-defined and being set to 0.0 in labels with no predicted samples. Use `zero_division` parameter to control this behavior.

from sklearn import tree
from sklearn.tree import DecisionTreeClassifier
dtree = DecisionTreeClassifier(random_state=0)
dtree_parameters = {'criterion': ['gini', 'entropy'],
                  'splitter': ['best', 'random'],
                  'max_depth': [1, 2, 3, 4, 5, 6, 7],
                  'max_features': [1, 2, 3, 4, 5, 6],
                  'min_samples_split':[5, 10, 15, 20, 25],
                  'min_samples_leaf':[1, 2, 3, 4, 5, 6]}
dtree_grid = GridSearchCV(dtree, dtree_parameters,cv=10)
dtree_grid.fit(X_train, y_train)

GridSearchCV(cv=10, estimator=DecisionTreeClassifier(random_state=0),
             param_grid={'criterion': ['gini', 'entropy'],
                         'max_depth': [1, 2, 3, 4, 5, 6, 7],
                         'max_features': [1, 2, 3, 4, 5, 6],
                         'min_samples_leaf': [1, 2, 3, 4, 5, 6],
                         'min_samples_split': [5, 10, 15, 20, 25],
                         'splitter': ['best', 'random']})

dtree_grid.best_params_

{'criterion': 'gini',
 'max_depth': 7,
 'max_features': 2,
 'min_samples_leaf': 3,
 'min_samples_split': 5,
 'splitter': 'random'}

dtree_test = DecisionTreeClassifier(random_state=0,max_depth=7 ,criterion='gini',max_features=2,min_samples_leaf=3,min_samples_split=5,splitter='random')
dtree_test.fit(X_train, y_train)

plt.figure(figsize = (20,8))
tree.plot_tree(dtree_test,precision=4,impurity=True,filled=True)
plt.show()

print('Train score: {:.4f}'.format(dtree_test.score(X_train, y_train)))
print('Test score: {:.4f}'.format(dtree_test.score(X_test, y_test)))

Train score: 0.9527
Test score: 0.9415

final_test_prediction = svm_test.predict(X_test)

print('Confusion Matrix \n', confusion_matrix(y_test, final_test_prediction))

Confusion Matrix 
 [[1385    0]
 [  84    0]]

print(classification_report(y_test, final_test_prediction))

              precision    recall  f1-score   support

           0       0.94      1.00      0.97      1385
           1       0.00      0.00      0.00        84

    accuracy                           0.94      1469
   macro avg       0.47      0.50      0.49      1469
weighted avg       0.89      0.94      0.92      1469

C:\Users\vochi\anaconda3\lib\site-packages\sklearn\metrics\_classification.py:1221: UndefinedMetricWarning:

Precision and F-score are ill-defined and being set to 0.0 in labels with no predicted samples. Use `zero_division` parameter to control this behavior.

print('Precision score: ', precision_score( y_test, final_test_prediction, average = 'macro'))

print('Recall Score: ', recall_score( y_test, final_test_prediction, average = 'macro'))

Precision score:  0.4714091218515997
Recall Score:  0.5

C:\Users\vochi\anaconda3\lib\site-packages\sklearn\metrics\_classification.py:1221: UndefinedMetricWarning:

Precision is ill-defined and being set to 0.0 in labels with no predicted samples. Use `zero_division` parameter to control this behavior.

• In reviewing the police shootings, we chose to look at the manner of death and whether or not the police officer used a taser, prior to resorting to their gun. This could imply they were trained to use restraint and hopefully prevent future incidents.
• The information could prove useful in the future, when implementing policies and procedures or designing trainings. The end result would be to decrease shooting incidents, while improving justice for the suspects.
• Since the data is historical and specific, we could not discover unstructured datapoints to improve on our understanding of all the factors. It would be interesting to see the officers’ training history and perhaps prior incidents and in the future monitoring of the vitals, to see whether they were in a heightened state prior to firing their weapon.
• We used precision and recall score to evaluate our classification report, since the dataset is biased with the majority of deaths showing no use of a taser. Unfortunately, without more sufficient information we cannot make any predictions on how police officers will respond, depending on the environment and their behavior.